Contingency management programs effectively treat many substance abusers. At the same time, typical contingency management programs are ineffective for many others. By definition, when contingency management fails it is a result of ineffective reinforcement of abstinence; in practice, we find these failures often occur when patients never contact the programmed consequences of abstinence. In many settings, shaping desired behavior by reinforcing successive approximations of the target behavior is a standard method for overcoming this problem. However, shaping has been infrequently examined in drug abuse treatment. In the past, such investigation was hampered by the lack of standardized methods for shaping behavior. Percentile schedules are a standardized, replicable, and effective means of shaping behavior; and thus, percentile schedules can be used to study shaping. This project uses percentile schedules to test the concept that effective shaping procedures will improve the outcomes seen with typical contingency management treatments, and that this improvement will result from shaping abstinence in those individuals unlikely to initiate periods of abstinence of sufficient length to earn the programmed reward. This concept is tested in two studies of smokers planning to quit. The initial study evaluates four versions of the percentile schedule shaping procedure to establish which best facilitates abstinence. The main study compares the outcomes of three groups. On one day prior to group assignment, participants are offered a financial incentive if they abstain from smoking. Participants are classified as hard-to-treat if they fail to earn this incentive, and group assignment is blocked on this variable. The first group receives contingency management treatment using a percentile schedule. The second receives standard contingency management treatment, reinforcing periods of abstinence sufficient to meet a strict fixed breath CO criterion. The third is a yoked control group receiving payments based on the earnings of subjects under the percentile schedule. This design allows the comparison of a shaping contingency management procedure to a standard procedure, and to the effects of non-contingent payments. These studies begin to address the critical problem of reducing the frequent failures of even the most effective treatments for substance abuse.